The Eye of HAL from 2001: A Space Odyssey
The vast ungainly "vessel" that is wireless communications is heading into uncharted waters and there is no going back. The continued rapid increase in complexity of 5G networks is leading to the slow but steady swapping out of old familiar Cisco router-style network balancing and efficiency paradigms with trained machine learning models that will inevitably seem closer to HAL of 2001: A Spacy Odyssey fame than to anything familiar in networking today.
Buzz words such as edge computing, 5G orchestration, cloud computing, and massive MIMO (multiple input multiple output) inspire excitement and enthusiasm in wireless circles. But, integrating this all together into wide ranging 5G networks capable of living up to the hype and expecations of consumers and businesses will require network management models more advanced than anything we have seen thus far. The human mind is simply not up to the task to program these new models. Instead, they will be trained.
Machine learning, or ML, is a sub field of artificial intelligence, or AI. Artificial neural networks are a sort of digitized representation of the brain's neural networks, as we understand them. Simply put, ML models are currently being trained up to understand and manage all the requirements of advanced 5G networks. As you may have guessed, this is no easy or streamlined undertaking. There are dozens of machine learning frameworks. Even selecting the most appropriate model to work with presents a host of issues. A framework suitable for voice recognition, such as the recurrent neural network, or RNN, may be wholly unsuited for encompassing the unique complexities of 5G networks.
Recurrent Neural Network Diagram
Once the ML framework is selected, there is the close to infathomable task of setting up the baseline attributes so that it is ready to train to learn everything it needs to learn to function. Some of the best and brightest minds are working on these tasks. A hive mind of unprecedented range and breadth must pull together for any chance of success. Unlike speech recognition models, which evolved over decades, 5G network models require accelerated development timelines. The finite lifespan of cellular network paradigms does not allow for the playful curiosity and research followed by rigorous testing and validation we have seen in other problems being solved by machine learning.
If the carriers are able to successfully bring ML to bear on their networks and improve them so that 5G's promised benefits can be widely realized, there are some easy to spot obstacles that will become apparent as soon as the models are deployed. Problems with diagnostics and troubleshooting come to mind first. In a conventional setting, a network engineer or Cisco expert could troubleshoot an entire network, isolate the problem and program the solution in the field, on the spot. It will not be as straightforward with ML-driven 5G networks. Neural networks are extremely complex, with even simple voice recognition apps designed to recognize no more than a few dozen command words, are levels of complexity higher than a fully integrated regional network. They will likely be closed black boxes of functionality, inaccessible to network administrators. Problem solving may frequently involve updating the ML model itself. Retraining it, using new data sets, new means of validation.
To remain relevant, carriers, local networks, private and public, and anyone else involved will need to be flexible and quickly grasp what can and what cannot be solved in the field, and what needs to get kicked back to the higher level of ML model development. This will not allow for the typical disconnect from the research origins seen when a service goes from idea, to prototype to consumer ready product.
Fleet Management Example
Soon after all this comes into place, local government will want to leverage the enormous potential that fully realized 5G networks can bring to its core services and functions. Amazon's fulfillment facilities, transportation fleets and cloud storage facilities are already managed to a significant degree by artificial intelligence. The benefits to efficiency and process improvement are direct and calculable.
A trained ML model managing your town's snow plowing schedule, recycling pickup, DPW fleet, permits and licenses, and other core public services sounds awesome, but also slightly dystopian, slightly eerie. However, the benefits will surely outweigh the downside, so being ready to adapt, move forward and benefit is in the best interest of your town.
As your expert telecommunications consultant, Hoplite Communications will continually stay abreast of key developments in 5G in order to help you determine the best ML-meets-5G solutions for your municipality, county, or other government agency.
That is, Hoplite can help optimize the physical appearance of 5G infrastructure in your community while, at the same time, ensuring that public services, your residents and businesses get the most out of 5G networks, Only Hoplite has the legal and technical expertise necessary to view your telecommunications and land use concerns in a holistic three dimensional manner.
Contact us today to find out more.